Re: [hybi] Adding multiplexing

Do firewalls and NATs handle SCTP properly?

While it would be nice to see HTTP become a layer in its own right, rather than only tied to TCP:
http://sat-net.com/L.Wood/dtn/http-dtn
http over SCTP is still pretty much experimental:
http://tools.ietf.org/html/draft-natarajan-http-over-sctp
expired draft.

and only the Uni of Delaware gang care.

L.
________________________________________
From: hybi-bounces@ietf.org [hybi-bounces@ietf.org] On Behalf Of Pieter Hintjens [ph@imatix.com]
Sent: 04 August 2010 13:25
To: Arman Djusupov
Cc: hybi@ietf.org
Subject: Re: [hybi] Adding multiplexing

On Wed, Aug 4, 2010 at 1:08 PM, Arman Djusupov <arman@noemax.com> wrote:

> Multiplexing requires just a channelID per frame and control frames for flow
> control and sub-channel management.

It also needs (if you want to specify interoperable multiplexing)
rules for clean tear down of channels, handling per-channel errors,
handling connection level errors, etc.  We did all this in AMQP, and
it gets *complex*.

If you don't care about interoperability and just want space for
experimentation, define a binary frame consisting of length + channel
ID, specify that channel IDs are experimental, and that channel ID
zero is for control frames.  Then define all higher frames types
inside that frame.

But I'd ask: why are we not looking at SCTP[1] as the underlying
protocol rather than TCP?  We'd get all this for free and SCTP has
other significant advantages too, such as multihoming.  SCTP is pretty
easy to use, TCP calls map to it.  Running HTTP over SCTP would give
these advantages to old applications as well.

The point: while multiplexing may be beneficial to some scenarios, it
is probably flawed to try to use WebSockets to gain this.  There are
IETF protocols that already do this, they have been around for a long
time, they have been proven to work better.[2]

-Pieter

[1] http://en.wikipedia.org/wiki/Stream_Control_Transmission_Protocol
[2] http://www.cis.udel.edu/~leighton/
_______________________________________________
hybi mailing list
hybi@ietf.org
https://www.ietf.org/mailman/listinfo/hybi